Think Visually Across Space and Time: Grounded Reasoning for Video Segmentation

Benchmark Model Rank Results
referring-video-object-segmentation-on-mevisTCoVP-M-4B (extra data)#1J&F: 65.2
referring-video-object-segmentation-on-reasonvosTCoVP-4B (extra data)#1J&F: 72.1
referring-video-object-segmentation-on-revosTCoVP-M-4B (extra data)#2J&F: 68.3